Most marketers using AI for email get generic AI slop. The kind of corporate-sounding "Subject: Don't miss out!" emails that nobody opens and nobody buys from.
The problem isn't AI. The problem is prompting.
I've been using ChatGPT, Claude, and other AI tools for email writing across SendFocus, my own brands (Outlaw Masks, The Hype Section), and during my contracted brand evangelist role at GetResponse. I've learned the difference between an AI-written email that converts and one that gets unsubscribed in 90 seconds is almost entirely about how you ask.
This is the playbook.
Looking for the strategic framework behind these prompts? See How to Use AI for Marketing in 2026 →
Watch the Full Breakdown
The 10-minute video walkthrough. Below: the full prompt library you can copy and paste.
Why Most AI-Written Emails Fail
Three reasons:
1. The prompt is too vague. "Write me a marketing email" gives you a generic template that sounds like every other AI-written email. Specificity in the prompt is what produces specificity in the output.
2. The strategy is missing. AI is good at writing. AI is bad at deciding what to write about. If you don't bring the strategy — who you're emailing, why, what action you want them to take, what proof you have — AI fills the gap with corporate-sounding filler.
3. The voice is unedited. Every AI tool has a default voice. It's polished, friendly, slightly bland. Without a forceful voice direction in your prompt, that's what you get. Your readers can smell it.
The fix for all three: a structured prompting approach.
The Framework: Strategy → AI → Edit → Test
Strategy. Before I open ChatGPT, I decide: who am I writing to, what do they need, what action do I want, and what proof or story do I have. This takes 60 seconds and is the highest-leverage minute in the whole process.
AI. I use a structured prompt (examples below) that feeds all the strategy into the model. The output is the first draft.
Edit. I edit for voice, brand consistency, and the specific details AI doesn't know. Usually 30-40% of the original output gets rewritten.
Test. Subject line A/B tests, send time experiments, segment-specific variations. AI helps me generate the variations faster, but the testing decisions are still mine.
The 5 ChatGPT Prompts I Actually Use
Copy them verbatim. Replace the bracketed placeholders with your specifics.
Prompt 1: Subject Line Generator
I need 10 subject line options for an email to [audience description].
Context:
- Product/service: [what you sell]
- The big benefit: [primary transformation or outcome]
- The specific hook: [a stat, story, or counterintuitive insight]
- The CTA: [what they should do next]
- Tone: [casual/professional/aggressive/playful — pick one]
- Maximum 50 characters per subject line
- Exclude emojis
- Avoid spam triggers: "FREE," "Act now," "Limited time," all caps
Give me a mix of:
- 3 curiosity-driven (open loop, question, or counterintuitive)
- 3 benefit-driven (specific outcome or transformation)
- 2 urgency-driven (time-sensitive or scarcity-based, but tasteful)
- 2 story-driven (personal or narrative hook)
For each subject line, write a 1-sentence explanation of why it works.Prompt 2: Welcome Sequence Writer
Write a 5-email welcome sequence for new subscribers to [brand name].
Subscriber context:
- How they joined: [lead magnet / purchase / referral / etc.]
- Their primary pain point: [what they need help with]
- Their desired outcome: [what success looks like]
- Their skepticism: [why they might not trust you yet]
Brand voice:
- [Describe your voice in 3-5 adjectives: e.g., direct, irreverent, generous, no-BS]
- Write at an 8th-grade reading level
- Use "you" and "I" liberally
- Short paragraphs (2-3 sentences max)
- No corporate jargon
Sequence structure:
Email 1 (immediate): Welcome + deliver the promise + set expectations
Email 2 (Day 2): Your origin story — why you care about this problem
Email 3 (Day 4): Your best actionable tip (give real value, no pitch)
Email 4 (Day 6): Social proof — case study or testimonial that mirrors their situation
Email 5 (Day 8): Soft introduction to your product/service + clear CTA
For each email, include:
- Subject line
- Preview text (first 90 characters)
- Full body copy
- One clear CTA
- P.S. line that adds a second hookPrompt 3: Sales / Promo Sequence Writer
Write a 3-email promotional sequence for [product/service name] priced at [price].
Audience:
- Who they are: [description]
- Their current state: [frustration or problem]
- Their desired state: [outcome they want]
- Why they haven't bought yet: [objection]
Offer details:
- What it is: [1-sentence description]
- Price: [amount]
- Deadline: [date or condition]
- Guarantee: [your risk reversal]
- Bonus (if any): [what they get extra]
Voice: [same voice guidelines as Prompt 2]
Sequence structure:
Email 1 (Day 1): The problem + the promise. No price. Focus on the transformation.
Email 2 (Day 3): The mechanism. How it works. Address the primary objection head-on.
Email 3 (Day 5): The close. Price, deadline, guarantee, bonus. Make the CTA impossible to miss.
Rules:
- Each email needs one clear CTA
- Use a story or specific example in at least one email
- Include one piece of proof (number, testimonial, or case study) per email
- No fake scarcity. If there's no real deadline, don't invent one.Prompt 4: Re-engagement Campaign
Write a 3-email re-engagement sequence for subscribers who haven't opened an email in [timeframe, e.g., 60 days].
Context:
- List size: [approximate number]
- Industry: [your niche]
- Typical open rate: [your average]
- What they originally signed up for: [lead magnet or promise]
Sequence goal: Either re-engage them or get them to unsubscribe voluntarily (clean the list).
Voice: [same as Prompt 2, but acknowledge the distance — don't pretend they haven't been ignoring you]
Email 1 (Day 1): "Are you still interested?" — direct, honest, low-pressure. Offer a one-click preference update.
Email 2 (Day 4): "Here's what you missed" — your best content, most popular resource, or biggest win since they went quiet.
Email 3 (Day 7): "This is the last one" — final attempt with a direct question about what they need. Clear unsubscribe link. No guilt.
Rules:
- Make unsubscribing feel like a clean choice, not a punishment
- Include a genuine value offer in Email 2 (don't just pitch)
- Track who engages — segment the re-engaged into a warmed-up sub-list
- Keep each email under 200 wordsPrompt 5: Weekly Newsletter / Content Email
Write a weekly newsletter email for [brand name].
This week's topic: [specific subject]
The core insight: [the one thing they should leave knowing]
The supporting proof: [data point, case study, or personal story]
The actionable takeaway: [what they should do differently this week]
Voice: [same as Prompt 2]
Structure:
- Hook (first 2-3 sentences that make them want to keep reading)
- The insight (the main idea, explained simply)
- The proof (story, data, or example that makes it real)
- The application (how to use this in their business/life this week)
- The CTA (one thing to click — article, resource, product, or reply)
- P.S. (a second hook, personal note, or teaser for next week)
Rules:
- One idea per email. Not two. Not five. One.
- 400-600 words max
- Use formatting for scanability: bold key phrases, short paragraphs, occasional single-sentence paragraphs
- The CTA should feel like the natural next step, not a hard pivot
- Preview text should complement (not repeat) the subject lineWhat AI Still Can't Do
Brand voice. AI defaults to friendly, polished, slightly corporate. If your brand voice is sharp, edgy, contrarian, or distinctly personal, AI will sand off the edges. You have to add them back.
Founder authenticity. AI can write about a founder's experience. It can't write as a founder with the specific details, anecdotes, and opinions that make founder-led email convert. The strongest emails I write start with my own draft and use AI to tighten — not the other way around.
Situational nuance. AI doesn't know what happened in your customer support inbox last week. It doesn't know your launch is delayed. It doesn't know your top customer just churned. The most valuable emails reference real, current, specific situations. You bring those.
Strategic decisions. AI will happily write an email for any strategy. It won't tell you the strategy is wrong. Knowing what to write — and what NOT to write — is still entirely your job.
The 80/20 of AI Email Writing
- 80% of the value comes from prompting better, not from using a better AI tool.
- The same ChatGPT account that produces generic slop for one marketer produces conversion-grade copy for another. The difference is the prompt.
- Strategy → AI → Edit → Test. Skip any step and you lose 30-50% of the quality.
- AI is a multiplier on your skill, not a replacement for it. If your strategy is weak, AI makes weak emails faster.
What This Looks Like in a Real Engagement
When I run a Strategy Day for a client, AI-powered email writing is one of the workstreams we build together. We pick the 5-7 prompts that match the client's business model, customize them for their voice and audience, and the client walks away with a copy-and-paste prompt library specific to their company. The goal: more email shipped every week, in your voice, without the blank-page tax.
If your business runs on email and you want AI to actually multiply your output instead of dilute it — book a Strategy Day. Four hours synchronous, a 25-page strategy report, 30 days of Slack support. $2,500 flat.
Want to see the SendFocus methodology before you book anything? Read the Revenue Engine, my free 32-page framework. Get the free Revenue Engine →
FAQ
Q: Does this work with Claude, Gemini, or other AI tools?
Yes. The prompts work across any modern AI model. ChatGPT, Claude, Gemini, Perplexity. The model matters less than the prompt structure.
Q: How do I prevent AI from sounding too corporate?
Voice rules at the bottom of every prompt. Explicit "avoid X" and "use Y" instructions. Then a manual edit pass where you read it out loud — if it sounds like a brand, not a human, rewrite.
Q: Should I disclose that AI helped write my emails?
Personal preference. The legal answer in most jurisdictions: no obligation. The ethical answer: if AI did 90%+ of the writing, mention it. If it did 30-50% and you heavily edited, treat it like spell-check — no disclosure needed.
Q: How fast does this actually make me?
For me, a welcome sequence that used to take 4-6 hours to write now takes 60-90 minutes (prompt → edit → polish). A newsletter issue that took 2-3 hours now takes 30-45 minutes. The speed is real. The quality stays the same — sometimes higher because I can iterate on more variations.
Related Reading
- How to Use AI for Marketing in 2026 (The Strategic Framework) →
- Email vs. Social Media in 2026: The Data Is In →
- Why Creators Are Leaving Social Media for Email in 2026 →
- Watch the full keynote: I'd Rather Have 1,000 Email Subscribers Than 100,000 Followers →
Stop guessing. Start sending emails that convert.
Book a Strategy Day and walk away with a custom prompt library, a 25-page strategy report, and 30 days of Slack support. $2,500 flat.
Reserve Your Strategy Day →